Gonzalo Muñoz

Joey Huchette, Gonzalo Muñoz, Thiago Serra et al.

In the past decade, deep learning became the prevalent methodology for predictive modeling thanks to the remarkable accuracy of deep neural networks in tasks such as computer vision and natural language processing. Meanwhile, the structure of neural networks converged back to simpler representations based on piecewise constant and piecewise linear functions such as the Rectified Linear Unit (ReLU), which became the most commonly used type of activation function in neural networks. That made certain types of network structure $\unicode{x2014}$such as the typical fully-connected feedforward neural network$\unicode{x2014}$ amenable to analysis through polyhedral theory and to the application of methodologies such as Linear Programming (LP) and Mixed-Integer Linear Programming (MILP) for a variety of purposes. In this paper, we survey the main topics emerging from this fast-paced area of work, which bring a fresh perspective to understanding neural networks in more detail as well as to applying linear optimization techniques to train, verify, and reduce the size of such networks.

Fabian Badilla, Marcos Goycoolea, Gonzalo Muñoz et al.

The use of Mixed-Integer Linear Programming (MILP) models to represent neural networks with Rectified Linear Unit (ReLU) activations has become increasingly widespread in the last decade. This has enabled the use of MILP technology to test-or stress-their behavior, to adversarially improve their training, and to embed them in optimization models leveraging their predictive power. Many of these MILP models rely on activation bounds. That is, bounds on the input values of each neuron. In this work, we explore the tradeoff between the tightness of these bounds and the computational effort of solving the resulting MILP models. We provide guidelines for implementing these models based on the impact of network structure, regularization, and rounding.

Pablo Carrasco, Gonzalo Muñoz

The non-convex nature of trained neural networks has created significant obstacles in their incorporation into optimization models. Considering the wide array of applications that this embedding has, the optimization and deep learning communities have dedicated significant efforts to the convexification of trained neural networks. Many approaches to date have considered obtaining convex relaxations for each non-linear activation in isolation, which poses limitations in the tightness of the relaxations. Anderson et al. (2020) strengthened these relaxations and provided a framework to obtain the convex hull of the graph of a piecewise linear convex activation composed with an affine function; this effectively convexifies activations such as the ReLU together with the affine transformation that precedes it. In this article, we contribute to this line of work by developing a recursive formula that yields a tight convexification for the composition of an activation with an affine function for a wide scope of activation functions, namely, convex or ``S-shaped". Our approach can be used to efficiently compute separating hyperplanes or determine that none exists in various settings, including non-polyhedral cases. We provide computational experiments to test the empirical benefits of these convex approximations.

Víctor Bucarey, Sophia Calderón, Gonzalo Muñoz et al.

Dealing with uncertainty in optimization parameters is an important and longstanding challenge. Typically, uncertain parameters are predicted accurately, and then a deterministic optimization problem is solved. However, the decisions produced by this so-called predict-then-optimize procedure can be highly sensitive to uncertain parameters. In this work, we contribute to recent efforts in producing decision-focused predictions, i.e., to build predictive models that are constructed with the goal of minimizing a regret measure on the decisions taken with them. We begin by formulating the exact expected regret minimization as a pessimistic bilevel optimization model. Then, we show computational complexity results of this problem, including its membership in NP. In combination with a known NP-hardness result, this establishes NP-completeness and discards its hardness in higher complexity classes. Using duality arguments, we reformulate it as a non-convex quadratic optimization problem. Finally, leveraging the quadratic reformulation, we show various computational techniques to achieve empirical tractability. We report extensive computational results on shortest-path and bipartite matching instances with uncertain cost vectors. Our results indicate that our approach can improve training performance over the approach of Elmachtoub and Grigas (2022), a state-of-the-art method for decision-focused learning.

Daniel Bienstock, Gonzalo Muñoz, Sebastian Pokutta

Deep learning has received much attention lately due to the impressive empirical performance achieved by training algorithms. Consequently, a need for a better theoretical understanding of these problems has become more evident in recent years. In this work, using a unified framework, we show that there exists a polyhedron which encodes simultaneously all possible deep neural network training problems that can arise from a given architecture, activation functions, loss function, and sample-size. Notably, the size of the polyhedral representation depends only linearly on the sample-size, and a better dependency on several other network parameters is unlikely (assuming $P\neq NP$). Additionally, we use our polyhedral representation to obtain new and better computational complexity results for training problems of well-known neural network architectures. Our results provide a new perspective on training problems through the lens of polyhedral theory and reveal a strong structure arising from these problems.